As weather patterns grow increasingly unpredictable, so too do crop pests. Pests are increasing in number, travelling to new places and resisting previous management strategies, challenging farmers in unprecedented ways. While pesticides offer a wholesale approach to controlling pests, farmers still lack the tools to quickly identify and manage the exact insects present in their fields. This unpredictability leaves farmers vulnerable to severe crop losses.
This situation is changing, however, thanks to the tools developed by Dr. Dirk Steinke and his team at the Centre for Biodiversity Genomics, in collaboration with Dr. Graham Taylor at the School of Engineering, with funding from Food from Thought.
The fundamental challenge being addressed, as Dirk puts it, is “how can we actually provide farmers with the information that [a certain pest] is present or not in a much shorter turnaround time than is currently the case?” Steinke’s team is addressing this challenge by developing tools that enable farmers to identify pests in near-real-time and manage them effectively without harming beneficial insects.
The solution begins with a low-cost, high-throughput imaging system that captures photos of up to 200 insect specimens per hour from field sampling. These images are then fed into an artificial intelligence (AI)-powered machine learning algorithm developed by Taylor, that can identify insects at the family level. This AI-powered identification is “often more reliable than our staff and they’ve been looking at insects, some of them for 10 years”, marvels Steinke.
These insects can also be screened through rapid DNA sequencing that provides species-level identification within a week, significantly shorter than traditional timeframes of weeks or months. The results of this DNA screening can then be cross-referenced with an extensive database that the team has compiled with comprehensive information from scientific papers about the pest species, their ecological roles, feeding behaviour, risk profiles and management practices.
Identifying the insects present in the field allows farmers not only to distinguish between pests and beneficial insects but also whether a pest is simply passing through the field without really causing damage. A single field may contain up to 100 different species of insects that are registered as pests in Canada, but, as Steinke notes, “[a] quarter of them are tree pests, so they must have been there [in the field] just passing by or were blown from the wind from nearby forests, so they’re not harmful to the crop.”
The current inability to quickly identify species can lead to blanket spraying of pesticides, which is costly when not needed and can also harm the beneficial insects present like pollinators. Instead, Steinke’s research will enable farmers to use targeted measures sparing the forest species and pollinators, transforming traditional pest management from educated guesswork to data-driven decision making.
The team is now developing a new feature that uses machine learning to identify newly discovered insects and predict whether they will indeed feed on the crops being grown, based on insect features and known relative species, extracted from imaging data using machine learning. This prediction of potential pest targets in advance, Steinke says, “gives you [farmers] a risk estimate of something that we don’t know”.
In addition to transforming farm pest management, Steinke’s team has also expanded on the governments lists of pest species present in Ontario. The team is working to get the additional species formally registered, so farmers have up-to-date information on what species are present.
While Steinke’s team has developed and validated these methods for managing pests, they are now looking to make them widely available through commercial or government support. In the coming years, these tools are poised to empower farmers by cutting out the guesswork and enabling them to respond to pest threats with greater confidence and effectiveness—critically so, as pest dynamics are getting increasingly unpredictable.
